I’m pleased to announce that our program PyTalk now works on its intended platform: the Raspberry Pi. To get the microphone to capture audio on the pi we needed to change the speech recognition library to use a sample rate of 44100 or 48000 instead of the default 16000. We also needed to use a chunk size of 512 and not the default 1024.
Our githrub repo now contains the necessary files to work on the pi. Although the installation instructions are simple, we will package the project on PyPi to make it even easier!
There are still some problems with capturing audio. The background noise and very little time delay is making it hard to distinguish between recognized words and gibberish. I’ve noticed some variables in the speech recognition library that will help with this, and I will play with these values later.
Overall what I really enjoyed about this sprint was the fact that not only did our project work, but we had it working very early, and as far as I know, we were the only team that had a working demo to show the class before the actual project due date!
Pharas is having some trouble getting the project packaged over at PyPi, but I think he will be able to get it up soon. Now to think about the next sprint… something with networking would be nice!